Thanks for the responses, it has been incredibly helpful. It seems the only sure-fire way of making a robust dynamic height control would be to send the pulses into a PLC and apply the THC there, delaying all signals equally. A PID loop could also be implemented in the PLC, and wouldn't be too difficult. The result would be a kick-a$$ dynamic height controller.
To me, it looks like there are too many barriers in Mach3 for this application (i.e. no velocity/acceleration control for THC, macros way, way too slow, applying offset directly to motor causes lost position, 1:1 buffer results in buffer under-runs, etc...). While I think it's likely possible, it doesn't look simple at all, while the PLC approach is straight forward and extremely robust.
Thanks,
Chris